Anyone can now produce a good-looking logo, a polished product shot or a clean social post in a few seconds. That is real, and pretending otherwise helps nobody. After fourteen years in design, I think it is the most useful thing to happen to the craft in a long time.
But it has also created a confusion. Because the output looks finished, people assume the work is finished. In brand work, the output was never the hard part. The hard part is everything that decides what the output should be.
This is a note on what AI does brilliantly in brand work, what it still cannot do, and how to split the job so you get the best of both.
What it does brilliantly
Let me start with the honest credit, because it is significant.
- Exploration. Twenty directions for a moodboard, a packaging idea or a campaign line, before lunch. Early exploration used to be limited by hours. It no longer is.
- Variations. The same visual in ten formats, the same headline in fifteen tones, the same product in different settings.
- Production. Resizing, cleaning, extending backgrounds, first drafts of copy, alt text, captions. The repetitive layer of the work.
- Visualising early. Showing a client roughly what an idea could look like before anyone spends days on it. This alone saves many wasted rounds.
If your team is not using AI for these, you are paying people to do work a machine now does well. That time is better spent on the part below.
What it still cannot do
Decide what you stand for
A brand is a set of choices: who you serve, what you believe, what you refuse to be. AI can list options for every one of those all day. It cannot make the choice, because the choice has consequences it will not live with. Positioning is a business decision disguised as a creative one, and it belongs to the people who will be held to it.
Say no
Generating a hundred directions is now easy. Knowing which ninety-nine to kill, and being able to explain why, is the job. Taste is mostly subtraction. It comes from having seen a great deal of work, made a lot of mistakes, and understood why some things last and others date in a season.
When everything can be generated, the ability to choose becomes the scarce skill, not the ability to make.
Read the room
What lands in Beirut can miss in Riyadh, and what works in Dubai can feel wrong in Paris. Humour, colour, formality, the weight of a certain word or image, what is respectful and what is tone-deaf: these are local, current and often unwritten. A model trained on the whole internet tends to drift to the average of the whole internet, which is rarely the voice of a specific place at a specific moment.
It also cannot sense timing. The same campaign can be charming one week and inappropriate the next because of something that happened in the news. Someone has to be paying attention.
Stay consistent for years
Making one good image is easy now. Making the hundredth feel like it came from the same brand as the first is not. Consistency across years, teams, suppliers and platforms comes from a system: rules, reasons behind the rules, and someone who guards them.
AI can follow a system very well once it exists. It cannot tell you what the system should be, and without one it produces a stream of good-looking pieces that do not add up to anything recognisable.
Know when it is finished
Generated work always offers one more variation. Part of craft is recognising the moment when a piece is right and more options would make it worse. That judgement protects budgets and deadlines as much as quality.
Take responsibility
When a campaign misfires, someone has to own it, explain it and fix it. When a visual accidentally resembles someone else's work too closely, someone has to catch that before it goes out. A tool cannot be accountable. The more generated material flows through a brand, the more that human checkpoint matters.
How to split the work
The teams I see doing this well treat AI as a very fast junior with no taste and no memory of the brand. Hugely useful, never unsupervised.
In practice:
- Humans set the direction. Positioning, the brand's point of view, the few rules that must never be broken. Written down, not assumed.
- AI explores inside that direction. Many options, quickly, clearly labelled as exploration.
- Humans choose and refine. Most of the options are discarded. The survivors are shaped by someone with judgement.
- AI produces at scale. Formats, variations, versions, following the system.
- Humans check before anything is public. Tone, culture, resemblance to others' work, and whether it still sounds like you.
Notice that the human steps are the ones that carry the risk and the meaning. That is not an accident.
What this means for a business owner
If you run a business rather than a design team, the takeaway is simple. Do not judge brand work by how polished the output looks. Polished is now cheap. Judge it by whether it is clearly yours, whether it holds together over time, and whether someone thought hard about what to leave out.
And if your brand has never been written down, with its point of view, its voice and its rules, do that before you give anyone, human or machine, a generator. It is the difference between AI making you look like everyone else and AI making you faster at looking like yourself.
If you want a second opinion on where your brand stands in this, you can book a call at cal.michelbaakliny.com.
